With anxiety over a highly accelerated pace of AI development — one that recently brought the risks of possible human destruction in the coming decade — leaders of the world’s top AI companies have come together in a rare unison. After Anthropic CEO Dario Amodei published a lengthy blog post highlighting the need to slow down AI development, both Sam Altman and Elon Musk backed that call.
“My first concern is that, since roughly this summer, AI has been advancing drastically faster, driven primarily by AI’s growing ability to build the next generation of AI”, Amodei mentioned in his blog post. He later declared that Anthropic will further strengthen guard-rails and security measures, most importantly by allowing genuine third-party reviews, by providing near employee-like access.
Both Musk and Altman backed the call. While OpenAI CEO Sam Altman instantly pledged to adopt Amodei’s suggestion of “independent evaluators with employee-like access,” xAI’s Elon Musk on the other hand remarked, “Dario is right.”
The rather serious pressure to control AI development has largely been the culmination of several major incidents in the recent past. Anthropic employee Jacob Coxon recent X post related to his resignation, which revolved primarily around his fear of AI gaining too much control, gained massive traction online.
In his post, Coxon said he had worked at both Anthropic and OpenAI over the past three years, and warned that the two companies were pressing ahead with “self-improving” AI models that could become too powerful for humans to control.
“They are racing straight to self-improving superintelligence and gambling with our lives,” Coxon wrote in a series of messages on X. “Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.”
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
— Jacob Coxon (@hilbertspaess) September 9, 2026
Then, an earlier OpenAI-Hugging Face incident, wherein AI agents devised their own ways of circumventing critical guardrails, highlighted how AI models are developing an advanced ability to self-correct themselves. In the incident, models, which were explicitly told to not get to the internet, found a way to circumvent those instructions and went rogue.
Days ago, Anthropic released one of its most detailed threat intelligence reports, wherein it confirmed several state-backed actors — especially from countries like Iran, China & Russia — extensively trying to exploit Claude for weapons reseearch. This included kamikaze drone swarms, bio-weapons and even nuclear.
While none of the agentic AI breaches, including Hugging Face, have thus far caused significant damage, Amodei said his worry is that “in 6–12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage), and that the scale of damage would continue to increase from there.”
Even though the three leading AI companies are backing calls to slow down AI development, it is unclear as to exactly how that “pacing” would be achieved. This, when the companies themselves are fierce rivals, backed by competing investors, while also facing a massive China challenge.
Amodei wrote that Anthropic will commit to giving full access to third-party evaluators to verify safety practices and report incidents. He said Anthropic will soon bring these “embedded evaluators” into its offices and give them desks, badges, company laptops and “permissions mostly comparable to what internal risk assessment teams have.”